An all-atom protein generative model.

Proteins mediate their functions through chemical interactions; modeling these interactions, which are typically through sidechains, is an important need in protein design. However, constructing an all-atom generative model requires an appropriate scheme for managing the jointly continuous and discrete nature of proteins encoded in the structure and sequence. We describe an all-atom diffusion model of protein structure, Protpardelle, which represents all sidechain states at once as a "superposition" state; superpositions defining a protein are collapsed into individual residue types and conformations during sample generation. When combined with sequence design methods, our model is able to codesign all-atom protein structure and sequence. Generated proteins are of good quality under the typical quality, diversity, and novelty metrics, and sidechains reproduce the chemical features and behavior of natural proteins. Finally, we explore the potential of our model to conduct all-atom protein design and scaffold functional motifs in a backbone- and rotamer-free way.

A cysteine-specific solubilizing tag strategy enables efficient chemical protein synthesis of difficult targets.

We developed a new cysteine-specific solubilizing tag strategy via a cysteine-conjugated succinimide. This solubilizing tag remains stable under common native chemical ligation conditions and can be efficiently removed with palladium-based catalysts. Utilizing this approach, we synthesized two proteins containing notably difficult peptide segments: interleukin-2 (IL-2) and insulin. This IL-2 chemical synthesis represents the simplest and most efficient approach to date, which is enabled by the cysteine-specific solubilizing tag to synthesize and ligate long peptide segments. Additionally, we synthesized a T8P insulin variant, previously identified in an infant with neonatal diabetes. We show that T8P insulin exhibits reduced bioactivity (a 30-fold decrease compared to standard insulin), potentially contributing to the onset of diabetes in these patients. In summary, our work provides an efficient tool to synthesize challenging proteins and opens new avenues for exploring research directions in understanding their biological functions.

A general platform for targeting MHC-II antigens via a single loop.

Class-II major histocompatibility complexes (MHC-IIs) are central to the communications between CD4+ T cells and antigen presenting cells (APCs), but intrinsic structural features associated with MHC-II make it difficult to develop a general targeting system with high affinity and antigen specificity. Here, we introduce a protein platform, Targeted Recognition of Antigen-MHC Complex Reporter for MHC-II (TRACeR-II), to enable the rapid development of peptide-specific MHC-II binders. TRACeR-II has a small helical bundle scaffold and uses an unconventional mechanism to recognize antigens via a single loop. This unique antigen-recognition mechanism renders this platform highly versatile and amenable to direct structural modeling of the interactions with the antigen. We demonstrate that TRACeR-II binders can be rapidly evolved across multiple alleles, while computational protein design can produce specific binding sequences for a SARS-CoV-2 peptide of unknown complex structure. TRACeR-II sheds light on a simple and straightforward approach to address the MHC peptide targeting challenge, without relying on combinatorial selection on complementarity determining region (CDR) loops. It presents a promising basis for further exploration in immune response modulation as well as a broad range of theragnostic applications.

Sparks of function by de novo protein design.

Information in proteins flows from sequence to structure to function, with each step causally driven by the preceding one. Protein design is founded on inverting this process: specify a desired function, design a structure executing this function, and find a sequence that folds into this structure. This 'central dogma' underlies nearly all de novo protein-design efforts. Our ability to accomplish these tasks depends on our understanding of protein folding and function and our ability to capture this understanding in computational methods. In recent years, deep learning-derived approaches for efficient and accurate structure modeling and enrichment of successful designs have enabled progression beyond the design of protein structures and towards the design of functional proteins. We examine these advances in the broader context of classical de novo protein design and consider implications for future challenges to come, including fundamental capabilities such as sequence and structure co-design and conformational control considering flexibility, and functional objectives such as antibody and enzyme design.

Photoswitchable binders enable temporal dissection of endogenous protein function.

General methods for spatiotemporal control of specific endogenous proteins would be broadly useful for probing protein function in living cells. Synthetic protein binders that bind and inhibit endogenous protein targets can be obtained from nanobodies, designed ankyrin repeat proteins (DARPins), and other small protein scaffolds, but generalizable methods to control their binding activity are lacking. Here, we report robust single-chain photoswitchable DARPins (psDARPins) for bidirectional optical control of endogenous proteins. We created topological variants of the DARPin scaffold by computer-aided design so fusion of photodissociable dimeric Dronpa (pdDronpa) results in occlusion of target binding at baseline. Cyan light induces pdDronpa dissociation to expose the binding surface (paratope), while violet light restores pdDronpa dimerization and paratope caging. Since the DARPin redesign leaves the paratope intact, the approach was easily applied to existing DARPins for GFP, ERK, and Ras, as demonstrated by relocalizing GFP-family proteins and inhibiting endogenous ERK and Ras with optical control. Finally, a Ras-targeted psDARPin was used to determine that, following EGF-activation of EGFR, Ras is required for sustained EGFR to ERK signaling. In summary, psDARPins provide a generalizable strategy for precise spatiotemporal dissection of endogenous protein function.

Fully synthetic platform to rapidly generate tetravalent bispecific nanobody-based immunoglobulins.

Nanobodies bind a target antigen with a kinetic profile similar to a conventional antibody, but exist as a single heavy chain domain that can be readily multimerized to engage antigen via multiple interactions. Presently, most nanobodies are produced by immunizing camelids; however, platforms for animal-free production are growing in popularity. Here, we describe the development of a fully synthetic nanobody library based on an engineered human VH3-23 variable gene and a multispecific antibody-like format designed for biparatopic target engagement. To validate our library, we selected nanobodies against the SARS-CoV-2 receptor-binding domain and employed an on-yeast epitope binning strategy to rapidly map the specificities of the selected nanobodies. We then generated antibody-like molecules by replacing the VH and VL domains of a conventional antibody with two different nanobodies, designed as a molecular clamp to engage the receptor-binding domain biparatopically. The resulting bispecific tetra-nanobody immunoglobulins neutralized diverse SARS-CoV-2 variants with potencies similar to antibodies isolated from convalescent donors. Subsequent biochemical analyses confirmed the accuracy of the on-yeast epitope binning and structures of both individual nanobodies, and a tetra-nanobody immunoglobulin revealed that the intended mode of interaction had been achieved. This overall workflow is applicable to nearly any protein target and provides a blueprint for a modular workflow for the development of multispecific molecules.

An all-atom protein generative model.

Proteins mediate their functions through chemical interactions; modeling these interactions, which are typically through sidechains, is an important need in protein design. However, constructing an all-atom generative model requires an appropriate scheme for managing the jointly continuous and discrete nature of proteins encoded in the structure and sequence. We describe an all-atom diffusion model of protein structure, Protpardelle, which instantiates a "superposition" over the possible sidechain states, and collapses it to conduct reverse diffusion for sample generation. When combined with sequence design methods, our model is able to co-design all-atom protein structure and sequence. Generated proteins are of good quality under the typical quality, diversity, and novelty metrics, and sidechains reproduce the chemical features and behavior of natural proteins. Finally, we explore the potential of our model conduct all-atom protein design and scaffold functional motifs in a backbone- and rotamer-free way.

Ig-VAE: Generative modeling of protein structure by direct 3D coordinate generation.

While deep learning models have seen increasing applications in protein science, few have been implemented for protein backbone generation-an important task in structure-based problems such as active site and interface design. We present a new approach to building class-specific backbones, using a variational auto-encoder to directly generate the 3D coordinates of immunoglobulins. Our model is torsion- and distance-aware, learns a high-resolution embedding of the dataset, and generates novel, high-quality structures compatible with existing design tools. We show that the Ig-VAE can be used with Rosetta to create a computational model of a SARS-CoV2-RBD binder via latent space sampling. We further demonstrate that the model's generative prior is a powerful tool for guiding computational protein design, motivating a new paradigm under which backbone design is solved as constrained optimization problem in the latent space of a generative model.

Interleukin-2 superkines by computational design.

Affinity maturation of protein­protein interactions is an important approach in the development of therapeutic proteins such as cytokines. Typical experimental strategies involve targeting the cytokine-receptor interface with combinatorial libraries and then selecting for higher-affinity variants. Mutations to the binding scaffold are usually not considered main drivers for improved affinity. Here we demonstrate that computational design can provide affinity-enhanced variants of interleukin-2 (IL-2) "out of the box" without any requirement for interface engineering. Using a strategy of global IL-2 structural stabilization targeting metastable regions of the three-dimensional structure, rather than the receptor binding interfaces, we computationally designed thermostable IL-2 variants with up to 40-fold higher affinity for IL-2Rß without any library-based optimization. These IL-2 analogs exhibited CD25-independent activities on T and natural killer (NK) cells both in vitro and in vivo, mimicking the properties of the IL-2 superkine "super-2" that was engineered through yeast surface display [A. M. Levin et al., Nature, 484, 529­533 (2012)]. Structure-guided stabilization of cytokines is a powerful approach to affinity maturation with applications to many cytokine and protein­protein interactions.

Protein sequence design with a learned potential.

The task of protein sequence design is central to nearly all rational protein engineering problems, and enormous effort has gone into the development of energy functions to guide design. Here, we investigate the capability of a deep neural network model to automate design of sequences onto protein backbones, having learned directly from crystal structure data and without any human-specified priors. The model generalizes to native topologies not seen during training, producing experimentally stable designs. We evaluate the generalizability of our method to a de novo TIM-barrel scaffold. The model produces novel sequences, and high-resolution crystal structures of two designs show excellent agreement with in silico models. Our findings demonstrate the tractability of an entirely learned method for protein sequence design.

Chimeric mutants of staphylococcal hemolysin, which act as both one-component and two-component hemolysin, created by grafting the stem domain.

Staphylococcus aureus expresses several hemolytic pore-forming toxins (PFTs), which are all commonly composed of three domains: cap, rim and stem. PFTs are expressed as soluble monomers and assemble to form a transmembrane ß-barrel pore in the erythrocyte cell membrane. The stem domain undergoes dramatic conformational changes to form a pore. Staphylococcal PFTs are classified into two groups: one-component α-hemolysin (α-HL) and two-component γ-hemolysin (γ-HL). The α-HL forms a homo-heptamer, whereas γ-HL is an octamer composed of F-component (LukF) and S-component (Hlg2). Because PFTs are used as materials for nanopore-based sensors, knowledge of the functional properties of PFTs is used to develop new, engineered PFTs. However, it remains challenging to design PFTs with a ß-barrel pore because their formation as transmembrane protein assemblies requires large conformational changes. In the present study, aiming to investigate the design principles of the ß-barrel formed as a consequence of the conformational change, chimeric mutants composed of the cap/rim domains of α-HL and the stem of LukF or Hlg2 were prepared. Biochemical characterization and electron microscopy showed that one of them assembles as a heptameric one-component PFT, whereas another participates as both a heptameric one- and heptameric/octameric two-component PFT. All chimeric mutants intrinsically assemble into SDS-resistant oligomers. Based on these observations, the role of the stem domain of these PFTs is discussed. These findings provide clues for the engineering of staphylococcal PFT ß-barrels for use in further promising applications.

De Novo Design of a Highly Stable Ovoid TIM Barrel: Unlocking Pocket Shape towards Functional Design.

The ability to finely control the structure of protein folds is an important prerequisite to functional protein design. The TIM barrel fold is an important target for these efforts as it is highly enriched for diverse functions in nature. Although a TIM barrel protein has been designed de novo, the ability to finely alter the curvature of the central beta barrel and the overall architecture of the fold remains elusive, limiting its utility for functional design. Here, we report the de novo design of a TIM barrel with ovoid (twofold) symmetry, drawing inspiration from natural beta and TIM barrels with ovoid curvature. We use an autoregressive backbone sampling strategy to implement our hypothesis for elongated barrel curvature, followed by an iterative enrichment sequence design protocol to obtain sequences which yield a high proportion of successfully folding designs. Designed sequences are highly stable and fold to the designed barrel curvature as determined by a 2.1 Å resolution crystal structure. The designs show robustness to drastic mutations, retaining high melting temperatures even when multiple charged residues are buried in the hydrophobic core or when the hydrophobic core is ablated to alanine. As a scaffold with a greater capacity for hosting diverse hydrogen bonding networks and installation of binding pockets or active sites, the ovoid TIM barrel represents a major step towards the de novo design of functional TIM barrels.

Structure-based protein design with deep learning.

Since the first revelation of proteins functioning as macromolecular machines through their three dimensional structures, researchers have been intrigued by the marvelous ways the biochemical processes are carried out by proteins. The aspiration to understand protein structures has fueled extensive efforts across different scientific disciplines. In recent years, it has been demonstrated that proteins with new functionality or shapes can be designed via structure-based modeling methods, and the design strategies have combined all available information - but largely piece-by-piece - from sequence derived statistics to the detailed atomic-level modeling of chemical interactions. Despite the significant progress, incorporating data-derived approaches through the use of deep learning methods can be a game changer. In this review, we summarize current progress, compare the arc of developing the deep learning approaches with the conventional methods, and describe the motivation and concepts behind current strategies that may lead to potential future opportunities.

Theoretical basis for stabilizing messenger RNA through secondary structure design.

RNA hydrolysis presents problems in manufacturing, long-term storage, world-wide delivery and in vivo stability of messenger RNA (mRNA)-based vaccines and therapeutics. A largely unexplored strategy to reduce mRNA hydrolysis is to redesign RNAs to form double-stranded regions, which are protected from in-line cleavage and enzymatic degradation, while coding for the same proteins. The amount of stabilization that this strategy can deliver and the most effective algorithmic approach to achieve stabilization remain poorly understood. Here, we present simple calculations for estimating RNA stability against hydrolysis, and a model that links the average unpaired probability of an mRNA, or AUP, to its overall hydrolysis rate. To characterize the stabilization achievable through structure design, we compare AUP optimization by conventional mRNA design methods to results from more computationally sophisticated algorithms and crowdsourcing through the OpenVaccine challenge on the Eterna platform. We find that rational design on Eterna and the more sophisticated algorithms lead to constructs with low AUP, which we term 'superfolder' mRNAs. These designs exhibit a wide diversity of sequence and structure features that may be desirable for translation, biophysical size, and immunogenicity. Furthermore, their folding is robust to temperature, computer modeling method, choice of flanking untranslated regions, and changes in target protein sequence, as illustrated by rapid redesign of superfolder mRNAs for B.1.351, P.1 and B.1.1.7 variants of the prefusion-stabilized SARS-CoV-2 spike protein. Increases in in vitro mRNA half-life by at least two-fold appear immediately achievable.

Optical control of fast and processive engineered myosins in vitro and in living cells.

Precision tools for spatiotemporal control of cytoskeletal motor function are needed to dissect fundamental biological processes ranging from intracellular transport to cell migration and division. Direct optical control of motor speed and direction is one promising approach, but it remains a challenge to engineer controllable motors with desirable properties such as the speed and processivity required for transport applications in living cells. Here, we develop engineered myosin motors that combine large optical modulation depths with high velocities, and create processive myosin motors with optically controllable directionality. We characterize the performance of the motors using in vitro motility assays, single-molecule tracking and live-cell imaging. Bidirectional processive motors move efficiently toward the tips of cellular protrusions in the presence of blue light, and can transport molecular cargo in cells. Robust gearshifting myosins will further enable programmable transport in contexts ranging from in vitro active matter reconstitutions to microfabricated systems that harness molecular propulsion.

Theoretical basis for stabilizing messenger RNA through secondary structure design.

RNA hydrolysis presents problems in manufacturing, long-term storage, world-wide delivery, and in vivo stability of messenger RNA (mRNA)-based vaccines and therapeutics. A largely unexplored strategy to reduce mRNA hydrolysis is to redesign RNAs to form double-stranded regions, which are protected from in-line cleavage and enzymatic degradation, while coding for the same proteins. The amount of stabilization that this strategy can deliver and the most effective algorithmic approach to achieve stabilization remain poorly understood. Here, we present simple calculations for estimating RNA stability against hydrolysis, and a model that links the average unpaired probability of an mRNA, or AUP, to its overall hydrolysis rate. To characterize the stabilization achievable through structure design, we compare AUP optimization by conventional mRNA design methods to results from more computationally sophisticated algorithms and crowdsourcing through the OpenVaccine challenge on the Eterna platform. These computational tests were carried out on both model mRNAs and COVID-19 mRNA vaccine candidates. We find that rational design on Eterna and the more sophisticated algorithms lead to constructs with low AUP, which we term 'superfolder' mRNAs. These designs exhibit wide diversity of sequence and structure features that may be desirable for translation, biophysical size, and immunogenicity, and their folding is robust to temperature, choice of flanking untranslated regions, and changes in target protein sequence, as illustrated by rapid redesign of superfolder mRNAs for B.1.351, P.1, and B.1.1.7 variants of the prefusion-stabilized SARS-CoV-2 spike protein. Increases in in vitro mRNA half-life by at least two-fold appear immediately achievable.

Identification of N-Terminally Diversified GLP-1R Agonists Using Saturation Mutagenesis and Chemical Design.

The glucagon-like peptide 1 receptor (GLP-1R) is a class B G-protein coupled receptor (GPCR) and diabetes drug target expressed mainly in pancreatic ß-cells that, when activated by its agonist glucagon-like peptide 1 (GLP-1) after a meal, stimulates insulin secretion and ß-cell survival and proliferation. The N-terminal region of GLP-1 interacts with membrane-proximal residues of GLP-1R, stabilizing its active conformation to trigger intracellular signaling. The best-studied agonist peptides, GLP-1 and exendin-4, share sequence homology at their N-terminal region; however, modifications that can be tolerated here are not fully understood. In this work, a functional screen of GLP-1 variants with randomized N-terminal domains reveals new GLP-1R agonists and uncovers a pattern whereby a negative charge is preferred at the third position in various sequence contexts. We further tested this sequence-structure-activity principle by synthesizing peptide analogues where this position was mutated to both canonical and noncanonical amino acids. We discovered a highly active GLP-1 analogue in which the native glutamate residue three positions from the N-terminus was replaced with the sulfo-containing amino acid cysteic acid (GLP-1-CYA). The receptor binding and downstream signaling properties elicited by GLP-1-CYA were similar to the wild type GLP-1 peptide. Computational modeling identified a likely mode of interaction of the negatively charged side chain in GLP-1-CYA with an arginine on GLP-1R. This work highlights a strategy of combinatorial peptide screening coupled with chemical exploration that could be used to generate novel agonists for other receptors with peptide ligands.

Tight and specific lanthanide binding in a de novo TIM barrel with a large internal cavity designed by symmetric domain fusion.

De novo protein design has succeeded in generating a large variety of globular proteins, but the construction of protein scaffolds with cavities that could accommodate large signaling molecules, cofactors, and substrates remains an outstanding challenge. The long, often flexible loops that form such cavities in many natural proteins are difficult to precisely program and thus challenging for computational protein design. Here we describe an alternative approach to this problem. We fused two stable proteins with C2 symmetry-a de novo designed dimeric ferredoxin fold and a de novo designed TIM barrel-such that their symmetry axes are aligned to create scaffolds with large cavities that can serve as binding pockets or enzymatic reaction chambers. The crystal structures of two such designs confirm the presence of a 420 cubic Ångström chamber defined by the top of the designed TIM barrel and the bottom of the ferredoxin dimer. We functionalized the scaffold by installing a metal-binding site consisting of four glutamate residues close to the symmetry axis. The protein binds lanthanide ions with very high affinity as demonstrated by tryptophan-enhanced terbium luminescence. This approach can be extended to other metals and cofactors, making this scaffold a modular platform for the design of binding proteins and biocatalysts.

HIV-1 VRC01 Germline-Targeting Immunogens Select Distinct Epitope-Specific B Cell Receptors.

Activating precursor B cell receptors of HIV-1 broadly neutralizing antibodies requires specifically designed immunogens. Here, we compared the abilities of three such germline-targeting immunogens against the VRC01-class receptors to activate the targeted B cells in transgenic mice expressing the germline VH of the VRC01 antibody but diverse mouse light chains. Immunogen-specific VRC01-like B cells were isolated at different time points after immunization, their VH and VL genes were sequenced, and the corresponding antibodies characterized. VRC01 B cell sub-populations with distinct cross-reactivity properties were activated by each immunogen, and these differences correlated with distinct biophysical and biochemical features of the germline-targeting immunogens. Our study indicates that the design of effective immunogens to activate B cell receptors leading to protective HIV-1 antibodies will require a better understanding of how the biophysical properties of the epitope and its surrounding surface on the germline-targeting immunogen influence its interaction with the available receptor variants in vivo.